• Strengthen the quality and specificity of peer feedback
• Build students' ability to identify essay strengths and weaknesses
• Transfer critical reading skills to students' own revision process
• Introduce responsible, purposeful academic use of GenAI

The Assignment

Students were asked to provide feedback on each other’s essay drafts during an in-class Peer Review Workshop. 

After reviewing a peer’s draft, students pasted their written feedback into Dartmouth Chat (using the GPT-5.2 Thinking model) alongside a structured prompt. The prompt instructed the model to act as an AI coach, flagging two places where the feedback was vague or unclear and explaining why and how it could be improved.

Students were then asked to revise their own feedback based on the AI's coaching—the model did not rewrite it for them. The activity was facilitated in-class with instructor guidance throughout.

What Worked

• Students received targeted, low-stakes feedback on the clarity and specificity of their feedback
• AI coaching surfaced revision opportunities without replacing student effort
• Clear guidelines and in-class scaffolding reduced student anxiety
• Experimental, hands-on structure resonated with learners
• Students recognized a new, academically interesting use of GenAI

Student Survey Results

Twelve students responded, rating four aspects of the GenAI integration on a five-point agreement scale. On the high end, 11 of 12 students agreed that guidelines for using AI were clear and resources to support use were available. On the low end, only five of 12 students would recommend adding more GenAI activities to the course.

Student Voices

It was very helpful to see how we could improve our skills in peer editing... It[the AI coach] gave me good feedback.

I thought it was interesting and helpful, and more practice could benefit and bring more from the AI, as it was like a trial run.

I liked that it was very experimental, so I was learning by doing.

Faculty Reflections: Building on What Worked

• Broaden the GenAI prompt to coach students on not just where feedback is vague but on more dimensions of where feedback could be improved
• Provide more context by training the model on the essay prompt for more tailored coaching
• Run the workshop earlier in the term—and possibly twice—for reinforcement
• Develop follow-up questions so students can engage more deeply with AI outputs
• Assess students' prior AI familiarity; consider trying Claude models alongside GPT models for model comparison